For both models, the nearest grid point to each rawinsonde site was determined and a vertical profile of temperature and relative humidity was extracted for valid times of 0000 and 1200 UTC. The differences between observations and model outputs were analyzed using several methods to assess the accuracy of the model soundings; biases and spreads were emphasized. For example, the data sets were broken into 10°C bins to examine the model accuracy in various temperature ranges. There is a high interest in model performance in the 0 to –20°C range, since this is where the vast majority of inflight icing conditions reside.
Biases and spreads in the temperature forecasts were less than those for the relative humidity data. In general, both the RUC and Alaska Eta models were warmer and drier than the observations, but to varying degrees depending on the temperature range and location. Geographic and temperature range biases were identified and will be discussed. In Alaska, two sounding sites were also found to have significant differences in representing cloudy conditions from other sites nearby. Some example cases from both datasets will be presented to better illustrate the differences.